dm: fix REQ_OP_ZONE_RESET bio handling
[linux/fpc-iii.git] / net / ipv6 / ipv6_sockglue.c
bloba531ba032b85da42d3b5eefcbd9e2e624b5a2868
1 /*
2 * IPv6 BSD socket options interface
3 * Linux INET6 implementation
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * Based on linux/net/ipv4/ip_sockglue.c
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
15 * FIXME: Make the setsockopt code POSIX compliant: That is
17 * o Truncate getsockopt returns
18 * o Return an optlen of the truncated length if need be
20 * Changes:
21 * David L Stevens <dlstevens@us.ibm.com>:
22 * - added multicast source filtering API for MLDv2
25 #include <linux/module.h>
26 #include <linux/capability.h>
27 #include <linux/errno.h>
28 #include <linux/types.h>
29 #include <linux/socket.h>
30 #include <linux/sockios.h>
31 #include <linux/net.h>
32 #include <linux/in6.h>
33 #include <linux/mroute6.h>
34 #include <linux/netdevice.h>
35 #include <linux/if_arp.h>
36 #include <linux/init.h>
37 #include <linux/sysctl.h>
38 #include <linux/netfilter.h>
39 #include <linux/slab.h>
41 #include <net/sock.h>
42 #include <net/snmp.h>
43 #include <net/ipv6.h>
44 #include <net/ndisc.h>
45 #include <net/protocol.h>
46 #include <net/transp_v6.h>
47 #include <net/ip6_route.h>
48 #include <net/addrconf.h>
49 #include <net/inet_common.h>
50 #include <net/tcp.h>
51 #include <net/udp.h>
52 #include <net/udplite.h>
53 #include <net/xfrm.h>
54 #include <net/compat.h>
55 #include <net/seg6.h>
57 #include <linux/uaccess.h>
59 struct ip6_ra_chain *ip6_ra_chain;
60 DEFINE_RWLOCK(ip6_ra_lock);
62 int ip6_ra_control(struct sock *sk, int sel)
64 struct ip6_ra_chain *ra, *new_ra, **rap;
66 /* RA packet may be delivered ONLY to IPPROTO_RAW socket */
67 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num != IPPROTO_RAW)
68 return -ENOPROTOOPT;
70 new_ra = (sel >= 0) ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL;
72 write_lock_bh(&ip6_ra_lock);
73 for (rap = &ip6_ra_chain; (ra = *rap) != NULL; rap = &ra->next) {
74 if (ra->sk == sk) {
75 if (sel >= 0) {
76 write_unlock_bh(&ip6_ra_lock);
77 kfree(new_ra);
78 return -EADDRINUSE;
81 *rap = ra->next;
82 write_unlock_bh(&ip6_ra_lock);
84 sock_put(sk);
85 kfree(ra);
86 return 0;
89 if (!new_ra) {
90 write_unlock_bh(&ip6_ra_lock);
91 return -ENOBUFS;
93 new_ra->sk = sk;
94 new_ra->sel = sel;
95 new_ra->next = ra;
96 *rap = new_ra;
97 sock_hold(sk);
98 write_unlock_bh(&ip6_ra_lock);
99 return 0;
102 struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
103 struct ipv6_txoptions *opt)
105 if (inet_sk(sk)->is_icsk) {
106 if (opt &&
107 !((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) &&
108 inet_sk(sk)->inet_daddr != LOOPBACK4_IPV6) {
109 struct inet_connection_sock *icsk = inet_csk(sk);
110 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
111 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie);
114 opt = xchg((__force struct ipv6_txoptions **)&inet6_sk(sk)->opt,
115 opt);
116 sk_dst_reset(sk);
118 return opt;
121 static bool setsockopt_needs_rtnl(int optname)
123 switch (optname) {
124 case IPV6_ADDRFORM:
125 case IPV6_ADD_MEMBERSHIP:
126 case IPV6_DROP_MEMBERSHIP:
127 case IPV6_JOIN_ANYCAST:
128 case IPV6_LEAVE_ANYCAST:
129 case MCAST_JOIN_GROUP:
130 case MCAST_LEAVE_GROUP:
131 case MCAST_JOIN_SOURCE_GROUP:
132 case MCAST_LEAVE_SOURCE_GROUP:
133 case MCAST_BLOCK_SOURCE:
134 case MCAST_UNBLOCK_SOURCE:
135 case MCAST_MSFILTER:
136 return true;
138 return false;
141 static int do_ipv6_setsockopt(struct sock *sk, int level, int optname,
142 char __user *optval, unsigned int optlen)
144 struct ipv6_pinfo *np = inet6_sk(sk);
145 struct net *net = sock_net(sk);
146 int val, valbool;
147 int retv = -ENOPROTOOPT;
148 bool needs_rtnl = setsockopt_needs_rtnl(optname);
150 if (!optval)
151 val = 0;
152 else {
153 if (optlen >= sizeof(int)) {
154 if (get_user(val, (int __user *) optval))
155 return -EFAULT;
156 } else
157 val = 0;
160 valbool = (val != 0);
162 if (ip6_mroute_opt(optname))
163 return ip6_mroute_setsockopt(sk, optname, optval, optlen);
165 if (needs_rtnl)
166 rtnl_lock();
167 lock_sock(sk);
169 switch (optname) {
171 case IPV6_ADDRFORM:
172 if (optlen < sizeof(int))
173 goto e_inval;
174 if (val == PF_INET) {
175 struct ipv6_txoptions *opt;
176 struct sk_buff *pktopt;
178 if (sk->sk_type == SOCK_RAW)
179 break;
181 if (sk->sk_protocol == IPPROTO_UDP ||
182 sk->sk_protocol == IPPROTO_UDPLITE) {
183 struct udp_sock *up = udp_sk(sk);
184 if (up->pending == AF_INET6) {
185 retv = -EBUSY;
186 break;
188 } else if (sk->sk_protocol != IPPROTO_TCP)
189 break;
191 if (sk->sk_state != TCP_ESTABLISHED) {
192 retv = -ENOTCONN;
193 break;
196 if (ipv6_only_sock(sk) ||
197 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) {
198 retv = -EADDRNOTAVAIL;
199 break;
202 fl6_free_socklist(sk);
203 __ipv6_sock_mc_close(sk);
206 * Sock is moving from IPv6 to IPv4 (sk_prot), so
207 * remove it from the refcnt debug socks count in the
208 * original family...
210 sk_refcnt_debug_dec(sk);
212 if (sk->sk_protocol == IPPROTO_TCP) {
213 struct inet_connection_sock *icsk = inet_csk(sk);
214 local_bh_disable();
215 sock_prot_inuse_add(net, sk->sk_prot, -1);
216 sock_prot_inuse_add(net, &tcp_prot, 1);
217 local_bh_enable();
218 sk->sk_prot = &tcp_prot;
219 icsk->icsk_af_ops = &ipv4_specific;
220 sk->sk_socket->ops = &inet_stream_ops;
221 sk->sk_family = PF_INET;
222 tcp_sync_mss(sk, icsk->icsk_pmtu_cookie);
223 } else {
224 struct proto *prot = &udp_prot;
226 if (sk->sk_protocol == IPPROTO_UDPLITE)
227 prot = &udplite_prot;
228 local_bh_disable();
229 sock_prot_inuse_add(net, sk->sk_prot, -1);
230 sock_prot_inuse_add(net, prot, 1);
231 local_bh_enable();
232 sk->sk_prot = prot;
233 sk->sk_socket->ops = &inet_dgram_ops;
234 sk->sk_family = PF_INET;
236 opt = xchg((__force struct ipv6_txoptions **)&np->opt,
237 NULL);
238 if (opt) {
239 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
240 txopt_put(opt);
242 pktopt = xchg(&np->pktoptions, NULL);
243 kfree_skb(pktopt);
245 sk->sk_destruct = inet_sock_destruct;
247 * ... and add it to the refcnt debug socks count
248 * in the new family. -acme
250 sk_refcnt_debug_inc(sk);
251 module_put(THIS_MODULE);
252 retv = 0;
253 break;
255 goto e_inval;
257 case IPV6_V6ONLY:
258 if (optlen < sizeof(int) ||
259 inet_sk(sk)->inet_num)
260 goto e_inval;
261 sk->sk_ipv6only = valbool;
262 retv = 0;
263 break;
265 case IPV6_RECVPKTINFO:
266 if (optlen < sizeof(int))
267 goto e_inval;
268 np->rxopt.bits.rxinfo = valbool;
269 retv = 0;
270 break;
272 case IPV6_2292PKTINFO:
273 if (optlen < sizeof(int))
274 goto e_inval;
275 np->rxopt.bits.rxoinfo = valbool;
276 retv = 0;
277 break;
279 case IPV6_RECVHOPLIMIT:
280 if (optlen < sizeof(int))
281 goto e_inval;
282 np->rxopt.bits.rxhlim = valbool;
283 retv = 0;
284 break;
286 case IPV6_2292HOPLIMIT:
287 if (optlen < sizeof(int))
288 goto e_inval;
289 np->rxopt.bits.rxohlim = valbool;
290 retv = 0;
291 break;
293 case IPV6_RECVRTHDR:
294 if (optlen < sizeof(int))
295 goto e_inval;
296 np->rxopt.bits.srcrt = valbool;
297 retv = 0;
298 break;
300 case IPV6_2292RTHDR:
301 if (optlen < sizeof(int))
302 goto e_inval;
303 np->rxopt.bits.osrcrt = valbool;
304 retv = 0;
305 break;
307 case IPV6_RECVHOPOPTS:
308 if (optlen < sizeof(int))
309 goto e_inval;
310 np->rxopt.bits.hopopts = valbool;
311 retv = 0;
312 break;
314 case IPV6_2292HOPOPTS:
315 if (optlen < sizeof(int))
316 goto e_inval;
317 np->rxopt.bits.ohopopts = valbool;
318 retv = 0;
319 break;
321 case IPV6_RECVDSTOPTS:
322 if (optlen < sizeof(int))
323 goto e_inval;
324 np->rxopt.bits.dstopts = valbool;
325 retv = 0;
326 break;
328 case IPV6_2292DSTOPTS:
329 if (optlen < sizeof(int))
330 goto e_inval;
331 np->rxopt.bits.odstopts = valbool;
332 retv = 0;
333 break;
335 case IPV6_TCLASS:
336 if (optlen < sizeof(int))
337 goto e_inval;
338 if (val < -1 || val > 0xff)
339 goto e_inval;
340 /* RFC 3542, 6.5: default traffic class of 0x0 */
341 if (val == -1)
342 val = 0;
343 np->tclass = val;
344 retv = 0;
345 break;
347 case IPV6_RECVTCLASS:
348 if (optlen < sizeof(int))
349 goto e_inval;
350 np->rxopt.bits.rxtclass = valbool;
351 retv = 0;
352 break;
354 case IPV6_FLOWINFO:
355 if (optlen < sizeof(int))
356 goto e_inval;
357 np->rxopt.bits.rxflow = valbool;
358 retv = 0;
359 break;
361 case IPV6_RECVPATHMTU:
362 if (optlen < sizeof(int))
363 goto e_inval;
364 np->rxopt.bits.rxpmtu = valbool;
365 retv = 0;
366 break;
368 case IPV6_TRANSPARENT:
369 if (valbool && !ns_capable(net->user_ns, CAP_NET_ADMIN) &&
370 !ns_capable(net->user_ns, CAP_NET_RAW)) {
371 retv = -EPERM;
372 break;
374 if (optlen < sizeof(int))
375 goto e_inval;
376 /* we don't have a separate transparent bit for IPV6 we use the one in the IPv4 socket */
377 inet_sk(sk)->transparent = valbool;
378 retv = 0;
379 break;
381 case IPV6_RECVORIGDSTADDR:
382 if (optlen < sizeof(int))
383 goto e_inval;
384 np->rxopt.bits.rxorigdstaddr = valbool;
385 retv = 0;
386 break;
388 case IPV6_HOPOPTS:
389 case IPV6_RTHDRDSTOPTS:
390 case IPV6_RTHDR:
391 case IPV6_DSTOPTS:
393 struct ipv6_txoptions *opt;
395 /* remove any sticky options header with a zero option
396 * length, per RFC3542.
398 if (optlen == 0)
399 optval = NULL;
400 else if (!optval)
401 goto e_inval;
402 else if (optlen < sizeof(struct ipv6_opt_hdr) ||
403 optlen & 0x7 || optlen > 8 * 255)
404 goto e_inval;
406 /* hop-by-hop / destination options are privileged option */
407 retv = -EPERM;
408 if (optname != IPV6_RTHDR && !ns_capable(net->user_ns, CAP_NET_RAW))
409 break;
411 opt = rcu_dereference_protected(np->opt,
412 lockdep_sock_is_held(sk));
413 opt = ipv6_renew_options(sk, opt, optname,
414 (struct ipv6_opt_hdr __user *)optval,
415 optlen);
416 if (IS_ERR(opt)) {
417 retv = PTR_ERR(opt);
418 break;
421 /* routing header option needs extra check */
422 retv = -EINVAL;
423 if (optname == IPV6_RTHDR && opt && opt->srcrt) {
424 struct ipv6_rt_hdr *rthdr = opt->srcrt;
425 switch (rthdr->type) {
426 #if IS_ENABLED(CONFIG_IPV6_MIP6)
427 case IPV6_SRCRT_TYPE_2:
428 if (rthdr->hdrlen != 2 ||
429 rthdr->segments_left != 1)
430 goto sticky_done;
432 break;
433 #endif
434 case IPV6_SRCRT_TYPE_4:
436 struct ipv6_sr_hdr *srh = (struct ipv6_sr_hdr *)
437 opt->srcrt;
439 if (!seg6_validate_srh(srh, optlen))
440 goto sticky_done;
441 break;
443 default:
444 goto sticky_done;
448 retv = 0;
449 opt = ipv6_update_options(sk, opt);
450 sticky_done:
451 if (opt) {
452 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
453 txopt_put(opt);
455 break;
458 case IPV6_PKTINFO:
460 struct in6_pktinfo pkt;
462 if (optlen == 0)
463 goto e_inval;
464 else if (optlen < sizeof(struct in6_pktinfo) || !optval)
465 goto e_inval;
467 if (copy_from_user(&pkt, optval, sizeof(struct in6_pktinfo))) {
468 retv = -EFAULT;
469 break;
471 if (sk->sk_bound_dev_if && pkt.ipi6_ifindex != sk->sk_bound_dev_if)
472 goto e_inval;
474 np->sticky_pktinfo.ipi6_ifindex = pkt.ipi6_ifindex;
475 np->sticky_pktinfo.ipi6_addr = pkt.ipi6_addr;
476 retv = 0;
477 break;
480 case IPV6_2292PKTOPTIONS:
482 struct ipv6_txoptions *opt = NULL;
483 struct msghdr msg;
484 struct flowi6 fl6;
485 struct sockcm_cookie sockc_junk;
486 struct ipcm6_cookie ipc6;
488 memset(&fl6, 0, sizeof(fl6));
489 fl6.flowi6_oif = sk->sk_bound_dev_if;
490 fl6.flowi6_mark = sk->sk_mark;
492 if (optlen == 0)
493 goto update;
495 /* 1K is probably excessive
496 * 1K is surely not enough, 2K per standard header is 16K.
498 retv = -EINVAL;
499 if (optlen > 64*1024)
500 break;
502 opt = sock_kmalloc(sk, sizeof(*opt) + optlen, GFP_KERNEL);
503 retv = -ENOBUFS;
504 if (!opt)
505 break;
507 memset(opt, 0, sizeof(*opt));
508 atomic_set(&opt->refcnt, 1);
509 opt->tot_len = sizeof(*opt) + optlen;
510 retv = -EFAULT;
511 if (copy_from_user(opt+1, optval, optlen))
512 goto done;
514 msg.msg_controllen = optlen;
515 msg.msg_control = (void *)(opt+1);
516 ipc6.opt = opt;
518 retv = ip6_datagram_send_ctl(net, sk, &msg, &fl6, &ipc6, &sockc_junk);
519 if (retv)
520 goto done;
521 update:
522 retv = 0;
523 opt = ipv6_update_options(sk, opt);
524 done:
525 if (opt) {
526 atomic_sub(opt->tot_len, &sk->sk_omem_alloc);
527 txopt_put(opt);
529 break;
531 case IPV6_UNICAST_HOPS:
532 if (optlen < sizeof(int))
533 goto e_inval;
534 if (val > 255 || val < -1)
535 goto e_inval;
536 np->hop_limit = val;
537 retv = 0;
538 break;
540 case IPV6_MULTICAST_HOPS:
541 if (sk->sk_type == SOCK_STREAM)
542 break;
543 if (optlen < sizeof(int))
544 goto e_inval;
545 if (val > 255 || val < -1)
546 goto e_inval;
547 np->mcast_hops = (val == -1 ? IPV6_DEFAULT_MCASTHOPS : val);
548 retv = 0;
549 break;
551 case IPV6_MULTICAST_LOOP:
552 if (optlen < sizeof(int))
553 goto e_inval;
554 if (val != valbool)
555 goto e_inval;
556 np->mc_loop = valbool;
557 retv = 0;
558 break;
560 case IPV6_UNICAST_IF:
562 struct net_device *dev = NULL;
563 int ifindex;
565 if (optlen != sizeof(int))
566 goto e_inval;
568 ifindex = (__force int)ntohl((__force __be32)val);
569 if (ifindex == 0) {
570 np->ucast_oif = 0;
571 retv = 0;
572 break;
575 dev = dev_get_by_index(net, ifindex);
576 retv = -EADDRNOTAVAIL;
577 if (!dev)
578 break;
579 dev_put(dev);
581 retv = -EINVAL;
582 if (sk->sk_bound_dev_if)
583 break;
585 np->ucast_oif = ifindex;
586 retv = 0;
587 break;
590 case IPV6_MULTICAST_IF:
591 if (sk->sk_type == SOCK_STREAM)
592 break;
593 if (optlen < sizeof(int))
594 goto e_inval;
596 if (val) {
597 struct net_device *dev;
598 int midx;
600 rcu_read_lock();
602 dev = dev_get_by_index_rcu(net, val);
603 if (!dev) {
604 rcu_read_unlock();
605 retv = -ENODEV;
606 break;
608 midx = l3mdev_master_ifindex_rcu(dev);
610 rcu_read_unlock();
612 if (sk->sk_bound_dev_if &&
613 sk->sk_bound_dev_if != val &&
614 (!midx || midx != sk->sk_bound_dev_if))
615 goto e_inval;
617 np->mcast_oif = val;
618 retv = 0;
619 break;
620 case IPV6_ADD_MEMBERSHIP:
621 case IPV6_DROP_MEMBERSHIP:
623 struct ipv6_mreq mreq;
625 if (optlen < sizeof(struct ipv6_mreq))
626 goto e_inval;
628 retv = -EPROTO;
629 if (inet_sk(sk)->is_icsk)
630 break;
632 retv = -EFAULT;
633 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
634 break;
636 if (optname == IPV6_ADD_MEMBERSHIP)
637 retv = ipv6_sock_mc_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
638 else
639 retv = ipv6_sock_mc_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_multiaddr);
640 break;
642 case IPV6_JOIN_ANYCAST:
643 case IPV6_LEAVE_ANYCAST:
645 struct ipv6_mreq mreq;
647 if (optlen < sizeof(struct ipv6_mreq))
648 goto e_inval;
650 retv = -EFAULT;
651 if (copy_from_user(&mreq, optval, sizeof(struct ipv6_mreq)))
652 break;
654 if (optname == IPV6_JOIN_ANYCAST)
655 retv = ipv6_sock_ac_join(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
656 else
657 retv = ipv6_sock_ac_drop(sk, mreq.ipv6mr_ifindex, &mreq.ipv6mr_acaddr);
658 break;
660 case MCAST_JOIN_GROUP:
661 case MCAST_LEAVE_GROUP:
663 struct group_req greq;
664 struct sockaddr_in6 *psin6;
666 if (optlen < sizeof(struct group_req))
667 goto e_inval;
669 retv = -EFAULT;
670 if (copy_from_user(&greq, optval, sizeof(struct group_req)))
671 break;
672 if (greq.gr_group.ss_family != AF_INET6) {
673 retv = -EADDRNOTAVAIL;
674 break;
676 psin6 = (struct sockaddr_in6 *)&greq.gr_group;
677 if (optname == MCAST_JOIN_GROUP)
678 retv = ipv6_sock_mc_join(sk, greq.gr_interface,
679 &psin6->sin6_addr);
680 else
681 retv = ipv6_sock_mc_drop(sk, greq.gr_interface,
682 &psin6->sin6_addr);
683 break;
685 case MCAST_JOIN_SOURCE_GROUP:
686 case MCAST_LEAVE_SOURCE_GROUP:
687 case MCAST_BLOCK_SOURCE:
688 case MCAST_UNBLOCK_SOURCE:
690 struct group_source_req greqs;
691 int omode, add;
693 if (optlen < sizeof(struct group_source_req))
694 goto e_inval;
695 if (copy_from_user(&greqs, optval, sizeof(greqs))) {
696 retv = -EFAULT;
697 break;
699 if (greqs.gsr_group.ss_family != AF_INET6 ||
700 greqs.gsr_source.ss_family != AF_INET6) {
701 retv = -EADDRNOTAVAIL;
702 break;
704 if (optname == MCAST_BLOCK_SOURCE) {
705 omode = MCAST_EXCLUDE;
706 add = 1;
707 } else if (optname == MCAST_UNBLOCK_SOURCE) {
708 omode = MCAST_EXCLUDE;
709 add = 0;
710 } else if (optname == MCAST_JOIN_SOURCE_GROUP) {
711 struct sockaddr_in6 *psin6;
713 psin6 = (struct sockaddr_in6 *)&greqs.gsr_group;
714 retv = ipv6_sock_mc_join(sk, greqs.gsr_interface,
715 &psin6->sin6_addr);
716 /* prior join w/ different source is ok */
717 if (retv && retv != -EADDRINUSE)
718 break;
719 omode = MCAST_INCLUDE;
720 add = 1;
721 } else /* MCAST_LEAVE_SOURCE_GROUP */ {
722 omode = MCAST_INCLUDE;
723 add = 0;
725 retv = ip6_mc_source(add, omode, sk, &greqs);
726 break;
728 case MCAST_MSFILTER:
730 struct group_filter *gsf;
732 if (optlen < GROUP_FILTER_SIZE(0))
733 goto e_inval;
734 if (optlen > sysctl_optmem_max) {
735 retv = -ENOBUFS;
736 break;
738 gsf = kmalloc(optlen, GFP_KERNEL);
739 if (!gsf) {
740 retv = -ENOBUFS;
741 break;
743 retv = -EFAULT;
744 if (copy_from_user(gsf, optval, optlen)) {
745 kfree(gsf);
746 break;
748 /* numsrc >= (4G-140)/128 overflow in 32 bits */
749 if (gsf->gf_numsrc >= 0x1ffffffU ||
750 gsf->gf_numsrc > sysctl_mld_max_msf) {
751 kfree(gsf);
752 retv = -ENOBUFS;
753 break;
755 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) {
756 kfree(gsf);
757 retv = -EINVAL;
758 break;
760 retv = ip6_mc_msfilter(sk, gsf);
761 kfree(gsf);
763 break;
765 case IPV6_ROUTER_ALERT:
766 if (optlen < sizeof(int))
767 goto e_inval;
768 retv = ip6_ra_control(sk, val);
769 break;
770 case IPV6_MTU_DISCOVER:
771 if (optlen < sizeof(int))
772 goto e_inval;
773 if (val < IPV6_PMTUDISC_DONT || val > IPV6_PMTUDISC_OMIT)
774 goto e_inval;
775 np->pmtudisc = val;
776 retv = 0;
777 break;
778 case IPV6_MTU:
779 if (optlen < sizeof(int))
780 goto e_inval;
781 if (val && val < IPV6_MIN_MTU)
782 goto e_inval;
783 np->frag_size = val;
784 retv = 0;
785 break;
786 case IPV6_RECVERR:
787 if (optlen < sizeof(int))
788 goto e_inval;
789 np->recverr = valbool;
790 if (!val)
791 skb_queue_purge(&sk->sk_error_queue);
792 retv = 0;
793 break;
794 case IPV6_FLOWINFO_SEND:
795 if (optlen < sizeof(int))
796 goto e_inval;
797 np->sndflow = valbool;
798 retv = 0;
799 break;
800 case IPV6_FLOWLABEL_MGR:
801 retv = ipv6_flowlabel_opt(sk, optval, optlen);
802 break;
803 case IPV6_IPSEC_POLICY:
804 case IPV6_XFRM_POLICY:
805 retv = -EPERM;
806 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
807 break;
808 retv = xfrm_user_policy(sk, optname, optval, optlen);
809 break;
811 case IPV6_ADDR_PREFERENCES:
813 unsigned int pref = 0;
814 unsigned int prefmask = ~0;
816 if (optlen < sizeof(int))
817 goto e_inval;
819 retv = -EINVAL;
821 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
822 switch (val & (IPV6_PREFER_SRC_PUBLIC|
823 IPV6_PREFER_SRC_TMP|
824 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
825 case IPV6_PREFER_SRC_PUBLIC:
826 pref |= IPV6_PREFER_SRC_PUBLIC;
827 break;
828 case IPV6_PREFER_SRC_TMP:
829 pref |= IPV6_PREFER_SRC_TMP;
830 break;
831 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
832 break;
833 case 0:
834 goto pref_skip_pubtmp;
835 default:
836 goto e_inval;
839 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC|
840 IPV6_PREFER_SRC_TMP);
841 pref_skip_pubtmp:
843 /* check HOME/COA conflicts */
844 switch (val & (IPV6_PREFER_SRC_HOME|IPV6_PREFER_SRC_COA)) {
845 case IPV6_PREFER_SRC_HOME:
846 break;
847 case IPV6_PREFER_SRC_COA:
848 pref |= IPV6_PREFER_SRC_COA;
849 case 0:
850 goto pref_skip_coa;
851 default:
852 goto e_inval;
855 prefmask &= ~IPV6_PREFER_SRC_COA;
856 pref_skip_coa:
858 /* check CGA/NONCGA conflicts */
859 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
860 case IPV6_PREFER_SRC_CGA:
861 case IPV6_PREFER_SRC_NONCGA:
862 case 0:
863 break;
864 default:
865 goto e_inval;
868 np->srcprefs = (np->srcprefs & prefmask) | pref;
869 retv = 0;
871 break;
873 case IPV6_MINHOPCOUNT:
874 if (optlen < sizeof(int))
875 goto e_inval;
876 if (val < 0 || val > 255)
877 goto e_inval;
878 np->min_hopcount = val;
879 retv = 0;
880 break;
881 case IPV6_DONTFRAG:
882 np->dontfrag = valbool;
883 retv = 0;
884 break;
885 case IPV6_AUTOFLOWLABEL:
886 np->autoflowlabel = valbool;
887 retv = 0;
888 break;
889 case IPV6_RECVFRAGSIZE:
890 np->rxopt.bits.recvfragsize = valbool;
891 retv = 0;
892 break;
895 release_sock(sk);
896 if (needs_rtnl)
897 rtnl_unlock();
899 return retv;
901 e_inval:
902 release_sock(sk);
903 if (needs_rtnl)
904 rtnl_unlock();
905 return -EINVAL;
908 int ipv6_setsockopt(struct sock *sk, int level, int optname,
909 char __user *optval, unsigned int optlen)
911 int err;
913 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
914 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
916 if (level != SOL_IPV6)
917 return -ENOPROTOOPT;
919 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
920 #ifdef CONFIG_NETFILTER
921 /* we need to exclude all possible ENOPROTOOPTs except default case */
922 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
923 optname != IPV6_XFRM_POLICY) {
924 lock_sock(sk);
925 err = nf_setsockopt(sk, PF_INET6, optname, optval,
926 optlen);
927 release_sock(sk);
929 #endif
930 return err;
932 EXPORT_SYMBOL(ipv6_setsockopt);
934 #ifdef CONFIG_COMPAT
935 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
936 char __user *optval, unsigned int optlen)
938 int err;
940 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
941 if (udp_prot.compat_setsockopt != NULL)
942 return udp_prot.compat_setsockopt(sk, level, optname,
943 optval, optlen);
944 return udp_prot.setsockopt(sk, level, optname, optval, optlen);
947 if (level != SOL_IPV6)
948 return -ENOPROTOOPT;
950 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER)
951 return compat_mc_setsockopt(sk, level, optname, optval, optlen,
952 ipv6_setsockopt);
954 err = do_ipv6_setsockopt(sk, level, optname, optval, optlen);
955 #ifdef CONFIG_NETFILTER
956 /* we need to exclude all possible ENOPROTOOPTs except default case */
957 if (err == -ENOPROTOOPT && optname != IPV6_IPSEC_POLICY &&
958 optname != IPV6_XFRM_POLICY) {
959 lock_sock(sk);
960 err = compat_nf_setsockopt(sk, PF_INET6, optname,
961 optval, optlen);
962 release_sock(sk);
964 #endif
965 return err;
967 EXPORT_SYMBOL(compat_ipv6_setsockopt);
968 #endif
970 static int ipv6_getsockopt_sticky(struct sock *sk, struct ipv6_txoptions *opt,
971 int optname, char __user *optval, int len)
973 struct ipv6_opt_hdr *hdr;
975 if (!opt)
976 return 0;
978 switch (optname) {
979 case IPV6_HOPOPTS:
980 hdr = opt->hopopt;
981 break;
982 case IPV6_RTHDRDSTOPTS:
983 hdr = opt->dst0opt;
984 break;
985 case IPV6_RTHDR:
986 hdr = (struct ipv6_opt_hdr *)opt->srcrt;
987 break;
988 case IPV6_DSTOPTS:
989 hdr = opt->dst1opt;
990 break;
991 default:
992 return -EINVAL; /* should not happen */
995 if (!hdr)
996 return 0;
998 len = min_t(unsigned int, len, ipv6_optlen(hdr));
999 if (copy_to_user(optval, hdr, len))
1000 return -EFAULT;
1001 return len;
1004 static int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1005 char __user *optval, int __user *optlen, unsigned int flags)
1007 struct ipv6_pinfo *np = inet6_sk(sk);
1008 int len;
1009 int val;
1011 if (ip6_mroute_opt(optname))
1012 return ip6_mroute_getsockopt(sk, optname, optval, optlen);
1014 if (get_user(len, optlen))
1015 return -EFAULT;
1016 switch (optname) {
1017 case IPV6_ADDRFORM:
1018 if (sk->sk_protocol != IPPROTO_UDP &&
1019 sk->sk_protocol != IPPROTO_UDPLITE &&
1020 sk->sk_protocol != IPPROTO_TCP)
1021 return -ENOPROTOOPT;
1022 if (sk->sk_state != TCP_ESTABLISHED)
1023 return -ENOTCONN;
1024 val = sk->sk_family;
1025 break;
1026 case MCAST_MSFILTER:
1028 struct group_filter gsf;
1029 int err;
1031 if (len < GROUP_FILTER_SIZE(0))
1032 return -EINVAL;
1033 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0)))
1034 return -EFAULT;
1035 if (gsf.gf_group.ss_family != AF_INET6)
1036 return -EADDRNOTAVAIL;
1037 lock_sock(sk);
1038 err = ip6_mc_msfget(sk, &gsf,
1039 (struct group_filter __user *)optval, optlen);
1040 release_sock(sk);
1041 return err;
1044 case IPV6_2292PKTOPTIONS:
1046 struct msghdr msg;
1047 struct sk_buff *skb;
1049 if (sk->sk_type != SOCK_STREAM)
1050 return -ENOPROTOOPT;
1052 msg.msg_control = optval;
1053 msg.msg_controllen = len;
1054 msg.msg_flags = flags;
1056 lock_sock(sk);
1057 skb = np->pktoptions;
1058 if (skb)
1059 ip6_datagram_recv_ctl(sk, &msg, skb);
1060 release_sock(sk);
1061 if (!skb) {
1062 if (np->rxopt.bits.rxinfo) {
1063 struct in6_pktinfo src_info;
1064 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1065 np->sticky_pktinfo.ipi6_ifindex;
1066 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr : np->sticky_pktinfo.ipi6_addr;
1067 put_cmsg(&msg, SOL_IPV6, IPV6_PKTINFO, sizeof(src_info), &src_info);
1069 if (np->rxopt.bits.rxhlim) {
1070 int hlim = np->mcast_hops;
1071 put_cmsg(&msg, SOL_IPV6, IPV6_HOPLIMIT, sizeof(hlim), &hlim);
1073 if (np->rxopt.bits.rxtclass) {
1074 int tclass = (int)ip6_tclass(np->rcv_flowinfo);
1076 put_cmsg(&msg, SOL_IPV6, IPV6_TCLASS, sizeof(tclass), &tclass);
1078 if (np->rxopt.bits.rxoinfo) {
1079 struct in6_pktinfo src_info;
1080 src_info.ipi6_ifindex = np->mcast_oif ? np->mcast_oif :
1081 np->sticky_pktinfo.ipi6_ifindex;
1082 src_info.ipi6_addr = np->mcast_oif ? sk->sk_v6_daddr :
1083 np->sticky_pktinfo.ipi6_addr;
1084 put_cmsg(&msg, SOL_IPV6, IPV6_2292PKTINFO, sizeof(src_info), &src_info);
1086 if (np->rxopt.bits.rxohlim) {
1087 int hlim = np->mcast_hops;
1088 put_cmsg(&msg, SOL_IPV6, IPV6_2292HOPLIMIT, sizeof(hlim), &hlim);
1090 if (np->rxopt.bits.rxflow) {
1091 __be32 flowinfo = np->rcv_flowinfo;
1093 put_cmsg(&msg, SOL_IPV6, IPV6_FLOWINFO, sizeof(flowinfo), &flowinfo);
1096 len -= msg.msg_controllen;
1097 return put_user(len, optlen);
1099 case IPV6_MTU:
1101 struct dst_entry *dst;
1103 val = 0;
1104 rcu_read_lock();
1105 dst = __sk_dst_get(sk);
1106 if (dst)
1107 val = dst_mtu(dst);
1108 rcu_read_unlock();
1109 if (!val)
1110 return -ENOTCONN;
1111 break;
1114 case IPV6_V6ONLY:
1115 val = sk->sk_ipv6only;
1116 break;
1118 case IPV6_RECVPKTINFO:
1119 val = np->rxopt.bits.rxinfo;
1120 break;
1122 case IPV6_2292PKTINFO:
1123 val = np->rxopt.bits.rxoinfo;
1124 break;
1126 case IPV6_RECVHOPLIMIT:
1127 val = np->rxopt.bits.rxhlim;
1128 break;
1130 case IPV6_2292HOPLIMIT:
1131 val = np->rxopt.bits.rxohlim;
1132 break;
1134 case IPV6_RECVRTHDR:
1135 val = np->rxopt.bits.srcrt;
1136 break;
1138 case IPV6_2292RTHDR:
1139 val = np->rxopt.bits.osrcrt;
1140 break;
1142 case IPV6_HOPOPTS:
1143 case IPV6_RTHDRDSTOPTS:
1144 case IPV6_RTHDR:
1145 case IPV6_DSTOPTS:
1147 struct ipv6_txoptions *opt;
1149 lock_sock(sk);
1150 opt = rcu_dereference_protected(np->opt,
1151 lockdep_sock_is_held(sk));
1152 len = ipv6_getsockopt_sticky(sk, opt, optname, optval, len);
1153 release_sock(sk);
1154 /* check if ipv6_getsockopt_sticky() returns err code */
1155 if (len < 0)
1156 return len;
1157 return put_user(len, optlen);
1160 case IPV6_RECVHOPOPTS:
1161 val = np->rxopt.bits.hopopts;
1162 break;
1164 case IPV6_2292HOPOPTS:
1165 val = np->rxopt.bits.ohopopts;
1166 break;
1168 case IPV6_RECVDSTOPTS:
1169 val = np->rxopt.bits.dstopts;
1170 break;
1172 case IPV6_2292DSTOPTS:
1173 val = np->rxopt.bits.odstopts;
1174 break;
1176 case IPV6_TCLASS:
1177 val = np->tclass;
1178 break;
1180 case IPV6_RECVTCLASS:
1181 val = np->rxopt.bits.rxtclass;
1182 break;
1184 case IPV6_FLOWINFO:
1185 val = np->rxopt.bits.rxflow;
1186 break;
1188 case IPV6_RECVPATHMTU:
1189 val = np->rxopt.bits.rxpmtu;
1190 break;
1192 case IPV6_PATHMTU:
1194 struct dst_entry *dst;
1195 struct ip6_mtuinfo mtuinfo;
1197 if (len < sizeof(mtuinfo))
1198 return -EINVAL;
1200 len = sizeof(mtuinfo);
1201 memset(&mtuinfo, 0, sizeof(mtuinfo));
1203 rcu_read_lock();
1204 dst = __sk_dst_get(sk);
1205 if (dst)
1206 mtuinfo.ip6m_mtu = dst_mtu(dst);
1207 rcu_read_unlock();
1208 if (!mtuinfo.ip6m_mtu)
1209 return -ENOTCONN;
1211 if (put_user(len, optlen))
1212 return -EFAULT;
1213 if (copy_to_user(optval, &mtuinfo, len))
1214 return -EFAULT;
1216 return 0;
1219 case IPV6_TRANSPARENT:
1220 val = inet_sk(sk)->transparent;
1221 break;
1223 case IPV6_RECVORIGDSTADDR:
1224 val = np->rxopt.bits.rxorigdstaddr;
1225 break;
1227 case IPV6_UNICAST_HOPS:
1228 case IPV6_MULTICAST_HOPS:
1230 struct dst_entry *dst;
1232 if (optname == IPV6_UNICAST_HOPS)
1233 val = np->hop_limit;
1234 else
1235 val = np->mcast_hops;
1237 if (val < 0) {
1238 rcu_read_lock();
1239 dst = __sk_dst_get(sk);
1240 if (dst)
1241 val = ip6_dst_hoplimit(dst);
1242 rcu_read_unlock();
1245 if (val < 0)
1246 val = sock_net(sk)->ipv6.devconf_all->hop_limit;
1247 break;
1250 case IPV6_MULTICAST_LOOP:
1251 val = np->mc_loop;
1252 break;
1254 case IPV6_MULTICAST_IF:
1255 val = np->mcast_oif;
1256 break;
1258 case IPV6_UNICAST_IF:
1259 val = (__force int)htonl((__u32) np->ucast_oif);
1260 break;
1262 case IPV6_MTU_DISCOVER:
1263 val = np->pmtudisc;
1264 break;
1266 case IPV6_RECVERR:
1267 val = np->recverr;
1268 break;
1270 case IPV6_FLOWINFO_SEND:
1271 val = np->sndflow;
1272 break;
1274 case IPV6_FLOWLABEL_MGR:
1276 struct in6_flowlabel_req freq;
1277 int flags;
1279 if (len < sizeof(freq))
1280 return -EINVAL;
1282 if (copy_from_user(&freq, optval, sizeof(freq)))
1283 return -EFAULT;
1285 if (freq.flr_action != IPV6_FL_A_GET)
1286 return -EINVAL;
1288 len = sizeof(freq);
1289 flags = freq.flr_flags;
1291 memset(&freq, 0, sizeof(freq));
1293 val = ipv6_flowlabel_opt_get(sk, &freq, flags);
1294 if (val < 0)
1295 return val;
1297 if (put_user(len, optlen))
1298 return -EFAULT;
1299 if (copy_to_user(optval, &freq, len))
1300 return -EFAULT;
1302 return 0;
1305 case IPV6_ADDR_PREFERENCES:
1306 val = 0;
1308 if (np->srcprefs & IPV6_PREFER_SRC_TMP)
1309 val |= IPV6_PREFER_SRC_TMP;
1310 else if (np->srcprefs & IPV6_PREFER_SRC_PUBLIC)
1311 val |= IPV6_PREFER_SRC_PUBLIC;
1312 else {
1313 /* XXX: should we return system default? */
1314 val |= IPV6_PREFER_SRC_PUBTMP_DEFAULT;
1317 if (np->srcprefs & IPV6_PREFER_SRC_COA)
1318 val |= IPV6_PREFER_SRC_COA;
1319 else
1320 val |= IPV6_PREFER_SRC_HOME;
1321 break;
1323 case IPV6_MINHOPCOUNT:
1324 val = np->min_hopcount;
1325 break;
1327 case IPV6_DONTFRAG:
1328 val = np->dontfrag;
1329 break;
1331 case IPV6_AUTOFLOWLABEL:
1332 val = np->autoflowlabel;
1333 break;
1335 case IPV6_RECVFRAGSIZE:
1336 val = np->rxopt.bits.recvfragsize;
1337 break;
1339 default:
1340 return -ENOPROTOOPT;
1342 len = min_t(unsigned int, sizeof(int), len);
1343 if (put_user(len, optlen))
1344 return -EFAULT;
1345 if (copy_to_user(optval, &val, len))
1346 return -EFAULT;
1347 return 0;
1350 int ipv6_getsockopt(struct sock *sk, int level, int optname,
1351 char __user *optval, int __user *optlen)
1353 int err;
1355 if (level == SOL_IP && sk->sk_type != SOCK_RAW)
1356 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1358 if (level != SOL_IPV6)
1359 return -ENOPROTOOPT;
1361 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen, 0);
1362 #ifdef CONFIG_NETFILTER
1363 /* we need to exclude all possible ENOPROTOOPTs except default case */
1364 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1365 int len;
1367 if (get_user(len, optlen))
1368 return -EFAULT;
1370 lock_sock(sk);
1371 err = nf_getsockopt(sk, PF_INET6, optname, optval,
1372 &len);
1373 release_sock(sk);
1374 if (err >= 0)
1375 err = put_user(len, optlen);
1377 #endif
1378 return err;
1380 EXPORT_SYMBOL(ipv6_getsockopt);
1382 #ifdef CONFIG_COMPAT
1383 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1384 char __user *optval, int __user *optlen)
1386 int err;
1388 if (level == SOL_IP && sk->sk_type != SOCK_RAW) {
1389 if (udp_prot.compat_getsockopt != NULL)
1390 return udp_prot.compat_getsockopt(sk, level, optname,
1391 optval, optlen);
1392 return udp_prot.getsockopt(sk, level, optname, optval, optlen);
1395 if (level != SOL_IPV6)
1396 return -ENOPROTOOPT;
1398 if (optname == MCAST_MSFILTER)
1399 return compat_mc_getsockopt(sk, level, optname, optval, optlen,
1400 ipv6_getsockopt);
1402 err = do_ipv6_getsockopt(sk, level, optname, optval, optlen,
1403 MSG_CMSG_COMPAT);
1404 #ifdef CONFIG_NETFILTER
1405 /* we need to exclude all possible ENOPROTOOPTs except default case */
1406 if (err == -ENOPROTOOPT && optname != IPV6_2292PKTOPTIONS) {
1407 int len;
1409 if (get_user(len, optlen))
1410 return -EFAULT;
1412 lock_sock(sk);
1413 err = compat_nf_getsockopt(sk, PF_INET6,
1414 optname, optval, &len);
1415 release_sock(sk);
1416 if (err >= 0)
1417 err = put_user(len, optlen);
1419 #endif
1420 return err;
1422 EXPORT_SYMBOL(compat_ipv6_getsockopt);
1423 #endif